I am tossing up the pros and cons of thread-based parallelism vs. process-based parallelism in terms of efficiency.
In both cases, I am binding to CPU cores and the given process/thread will be sole owner of the core.
Given these constraints, I was thinking that efficiency of both would be about the same, save for TLB usage where the multi-threaded solution would have the efficiency advantage.
Does this seem right?